Review
Landscape genetics of plants
Rolf Holderegger1,2, Dominique Buehler1,2, Felix Gugerli1 and Stéphanie Manel3,4
1
WSL Swiss Federal Research Institute, Zürcherstrasse 111, CH-8903 Birmensdorf, Switzerland
2
ETH Zürich, Department of Environmental Sciences, Universitätsstrasse 16, CH-8092 Zürich, Switzerland
3
Laboratoire Population Environnement Développement, UMR 151 UP/IRD, Université de Provence, 3 place Victor Hugo,
13331 Marseille Cedex 03, France
4
Laboratoire d’Écologie Alpine (LECA), CNRS UMR 5553, Université Joseph Fourier, BP 53, 2233 Rue de la Piscine,
38041 Grenoble Cedex 09, France
Landscape genetics is the amalgamation of landscape
ecology and population genetics to help with under-
standing microevolutionary processes such as gene flow
and adaptation. In this review, we examine why land-
scape genetics of plants lags behind that of animals,
both in number of studies and consideration of land-
scape elements. The classical landscape distance/resis-
tance approach to study gene flow is challenging in
plants, whereas boundary detection and the assessment
of contemporary gene flow are more feasible. By con-
trast, the new field of landscape genetics of adaptive
genetic variation, establishing the relationship between
adaptive genomic regions and environmental factors in
natural populations, is prominent in plant studies. Land-
scape genetics is ideally suited to study processes such
as migration and adaptation under global change.
The booming field of landscape genetics
Landscape genetics is the amalgamation of landscape
ecology and population genetics to help with understand-
ing microevolutionary processes such as gene flow and
adaptation on the scale of natural landscapes [1]. This
field investigates how landscape elements and environ-
mental factors influence the spatial distribution of genetic
variation. For instance, landscape genetics assesses how
landscape elements such as forests or open fields affect
gene flow (see Glossary) in species inhabiting semi-natural
habitat remnants in an otherwise unsuitable and inten-
sively used landscape; a question of interest in conserva-
tion management. Similarly, landscape genetics analyzes
whether non-crop strips provide effective barriers to gene
flow between organic and genetically modified crops; a
question of practical importance in agriculture. Landscape
genetics also deals with how environmental factors such as
temperature or precipitation affect adaptive genetic varia-
tion; relevant information in the context of climate change
[2,3]. Landscape genetics still suffers from a lack of theo-
retical foundations and expectations [4–6], but empirical
studies are typically characterized by including geo-refer-
enced individuals or populations genotyped at multiple loci
and at least one landscape or environmental variable of
interest (in addition to geographical distance) measured at
or in between sampling locations [7]. Depending on the
landscape or environmental variables assessed, the study
area might be large (thousands of km2), for example, when
studying the adaptive response of a dominant forest tree to
Corresponding author: Holderegger, R. (rolf.holderegger@wsl.ch).
1360-1385/$ – see front matter ß 2010 Elsevier Ltd. All rights reserved. doi:10.1016/j.tplants.2010.09.002 Trends in Plant Science, December 2010, Vol. 15, No. 12
Glossary
AFLPs: amplified fragment length polymorphisms. Dominant and anonymous
DNA fingerprints.
Allele distribution model: a (usually non-spatial) statistical description of how
allele frequencies at loci linked to genes under selection are influenced by
distinct environmental factors.
Assignment test: statistical approach that assigns an individual to that
sampled population from which its multilocus genotype is most likely to be
derived.
Bayesian inference: statistical approach using prior data or information to
estimate posterior probabilities of a hypothesis to be correct. For instance,
prior information (or just a guess) on migration rates can be used in Bayesian
assignment tests to infer contemporary gene flow.
Chord distance (dc): a measure of genetic similarity between individuals or
populations based on allele frequencies and located on a sphere. Taking values
between 0 and 1.
Gene flow: exchange of genes among populations or individuals.
Genome scan: genotyping of many samples at a large number of (potentially
anonymous) molecular markers across the genome, used in outlier detection
and the landscape genetic analysis of adaptive genetic variation.
Hardy-Weinberg equilibrium: population genetic principle stating that allele
and genotype frequencies reach equilibrium and stay constant in random
mating populations assuming large population size, no selection, no migration
and no mutation.
Isolation by distance: spatial pattern describing decreasing genetic relatedness
of populations or individuals with increasing geographic distance.
Kriging: geostatistical technique to interpolate the value of a parameter of
interest at an unobserved geographic location from observations of this value
at nearby locations.
Landscape: an area spatially heterogeneous in one or more biotic and abiotic
factors of interest. From the human perspective, a landscape is perceived as a
kilometers-wide environmental mosaic.
Landscape distance: distance-like measurements parameterizing the landscape
between two localities, for example, geographical distance along a river,
number of roads to be crossed, or percentage of forest cover in a corridor strip
connecting two localities.
Landscape resistance: permeability values of different landscape elements,
describing their resistance to migration and dispersal. From these resistance
values, different types of cumulative resistances between localities can be
calculated, for example, least cost paths.
Least cost path: length of a path minimizing the cumulative landscape
resistance between two localities.
Mantel test: a permutation-based statistical test describing the correlation
between two distance matrices. A partial Mantel simultaneously accounts for
the effects of a third (or several) distance matrix (matrices).
Monmonier’s algorithm: detects genetic boundaries by finding the path
exhibiting the largest genetic distances among neighboring populations.
Nei’s genetic distance (dn): measure of genetic similarity based on the
probability that two randomly chosen alleles from different populations or
individuals are identical. Taking values between 0 and 1.
Neutral molecular marker: molecular markers not affected by natural selection.
PCNMs: principal coordinates of neighbor matrices are a spectral representa-
tion of all spatial relationships among sampling locations and describe all
spatial scales that can be accommodated by the sampling design. PCNMs are
calculated from principal coordinate analysis (PCoA). In landscape genetics,
they are used to account for spatial relationships among sampling locations
and for unaccounted environmental factors.
Permutation test: type of statistical tests that rely on resampling of data for
significance testing and not on theoretical probability distributions as in
classical statistics. For instance, permutation tests are used to account for the
non-independence of genetic, geographic and landscape distances among
sampling locations in landscape genetics.
675
Review
Population differentiation (FST): different measurements of the amount of
genetic variation found between populations. Most often used is Wright’s FST,
taking values between 0 and 1. Similar measurements can be calculated for
genetic differentiation among groups of populations (FCT).
SNPs: single-nucleotide polymorphisms. Bi-allelic, co-dominant molecular
markers of known position in the genome. Increasingly used in landscape
genetics.
Wombling: approach to search for areas across an interpolated allele
frequency surface where the slopes of the surface are particularly steep. Used
to infer genetic breaks or discontinuities.
latitude, or small (hundreds of m2), for example, when
assessing pollen dispersal in a forest herb.
Given its appeal to both basic and applied sciences,
landscape genetics has received increasing attention in
recent years. The field currently deals with two main
topics. First, landscape and environment are evaluated
by considering their effects on migration, dispersal and
gene flow, which are measured in terms of neutral genetic
variation [7]. Second, landscape genetics has started to
explore the interaction between environment and adaptive
genetic variation in natural populations and individuals, a
new field often referred to as landscape genomics [8–10].
Landscape genetics of plants
A recent survey [7] showed that the majority of landscape
genetic studies have dealt with animals and only seldom
with plants (18%). Furthermore, studies on animals and
plants differ in study design and analytical approaches.
Why are there such differences in landscape genetic stud-
ies on plants and animals?
When dealing with gene flow, landscape genetics con-
siders the landscape between sampling locations. When
studying adaptive genetic variation, landscape genetics
deals with the particular environment at sampling loca-
tions [11]. These two situations make plants either less or
more amenable to landscape genetic analysis than ani-
mals. Because sessile plants directly respond to the envi-
ronment at their growing site, the study of adaptive genetic
variation in plants is straightforward. By contrast, corre-
sponding animal studies must account for the environment
of the entire home range or of all resource sites of indivi-
duals or populations. However, when assessing migration,
dispersal and gene flow, studying plants is more complex
than investigating animals. In animals, individuals dis-
perse to other locations and provide gene flow when mat-
ing. In vascular plants, gene flow mainly happens through
two processes, the dispersal of diploid embryos in seeds
and of haploid male gametes in pollen. In wind-pollinated
and -dispersed plants, an abiotic factor acts as the primary
pollen and seed vector. In insect-pollinated and animal-
dispersed plants, animals act as dispersal vectors of pro-
pagules. Here, it is a moving animal that reacts to the
landscape and not the plant itself. Landscape genetic
studies investigating gene flow in plants thus deal with
the problem that two propagule types are dispersed by
particular vectors. For instance, the fragmentation effect of
roads on plants is not direct, but induced by the indirect
effects of roads on pollinators or animal seed dispersers,
whereas wind-pollinated and -dispersed plants might not
be affected by roads [12].
Published reviews on landscape genetics have focused
on animals, and many plant studies have not included
676
Trends in Plant Science Vol.15 No.12
landscape elements, apart from geographical distance.
Given the characteristics of plants and the present shortage
of empirical studies, it seems relevant to provide an over-
view focusing on plants and stressing associated benefits
and limitations of common landscape genetic approaches. In
this review, we will first summarize and discuss the main
approaches currently used to study gene flow at the land-
scape scale: the landscape distance/resistance approach, the
overlay technique and the assessment of contemporary gene
flow. We will then examine landscape genetic approaches
exploring adaptively relevant genetic variation [13,14].
Gene flow on the landscape scale
Landscape distance/resistance
This classical approach correlates a matrix of genetic dis-
tances as indirect measurements of gene flow [15] with
matrices of landscape distances/resistances and geograph-
ic distances (Figure 1a). The genetic matrix consists of
pairwise genetic distances among all pairs of individuals or
populations studied. Various estimators can be used for
this purpose, for example, genetic Chord distance (dc),
Nei’s distance (dn) or population differentiation (FST)
[7,16]. The geographic distance matrix contains the
straight line distances among all sampling locations. By
contrast, the landscape distance matrix varies and can
contain the length, area or percentage of cover of landscape
elements, such as ditches, wetlands or woodlands in corri-
dors of a certain width between pairs of sampling locations.
It can also be a 0/1 matrix, for example, when some
sampling locations are separated by a river and others
are not. In a more sophisticated design, land cover/land use
and topography are taken from existing geographic infor-
mation system (GIS) data or from field surveys, and a level
of resistance to movement is given to each raster cell
containing a particular landscape feature. For instance,
forests might hinder gene flow whereas open fields have no
effect, and a high resistance is thus assigned to forested
grid cells, whereas open-field cells receive low resistance
values. Using GIS technology, the length of the shortest
path connecting two sampling locations is determined by
maximizing movement through low-resistance cells [17].
Such least-cost paths form the entries for a landscape
resistance matrix. Resistance assignment relies on expert
knowledge or a priori ecological information on study
organisms [18], and several alternative landscape resis-
tance models are generally tested per study.
For statistical analysis, the genetic distance matrix is
either correlated to the geographic distance and landscape
distance/resistance matrices separately using a Mantel
test, or the effect of geographic distance is first partialed
out before estimating correlation with landscape distance/
resistance in a partial Mantel test [19] (Figure 1a). This
procedure enables the effects of geographic distance (i.e.
isolation by distance [20]) and landscape elements on gene
flow to be disentangled [21]. However, partial Mantel tests
have been criticized because of their permutation proce-
dure (e.g. [22]), and a variety of alternative approaches
have been suggested [23,24].
The landscape resistance approach and especially least
cost path analysis have been popular in animal studies, but
there are few such studies in plants [7] (but see [25,26]).
(Figure_1)TD$IG][ Review Trends in Plant Science Vol.15 No.12

(a)

[ Genetic
distance ] X
[ Geographic
distance ]
[ Landscape
distance ]
F ST, d c, (Partial) Landscape features
d n, etc. Mantel test Least-cost path

(b)

Bayesian clustering Genetic discontinuity Interpolation

Barrier detection kriging

Spatial coincidence, boundary overlap statistics

(c)
[ Contemporary
gene flow ] X [ Landscape
distance ]
Parentage analysis Landscape
assignment tests features

Multiple regression
(permutation testing)

TRENDS in Plant Science

Figure 1. Schematic summary of the three major analytical approaches that are currently used by landscape geneticists to study gene flow. (a–c) Individuals or populations
(circles) are studied in a simple landscape consisting of two land cover types: meadows (open) and forests (hatched green). (a) In the landscape distance/resistance
approach, a matrix of genetic distances between all pairs of individuals (e.g. Nei’s genetic distance dn, Chord distance dc [17]) or between all pairs of populations (e.g.
genetic differentiation FST [16]) is correlated with a matrix of geographic distances between sampling locations (solid red straight lines) and landscape distances (landscape
features such as percentage of forest between sampling locations) or landscape resistance (e.g. length of least-cost path [17] assuming a high resistance value of forests to
gene flow; broken red lines) in Mantel or partial Mantel tests [7]. (b) The overlay technique uses several methods to cluster individuals into groups (e.g. Bayesian clustering;
groups indicated by open and red circles [5]), to detect genetic discontinuities or barriers (thick red line [36]) or to interpolate genetic distances among individuals or
populations (e.g. kriging resulting in genetic isolines given in red [1]). These genetic groups, barriers, genetic discontinuities or isolines are overlaid on topographical or
land cover/land use maps to search for spatial coincidences of these genetic structures with landscape elements. (c) Contemporary gene flow events or migrants (red
arrows indicating direction and abundance) can be assessed by either parentage analysis (paternity or maternity analysis [58]) or assignment tests [36]. Any number of
matrices of landscape distances between sampling locations connected by migration or gene flow is correlated with a matrix of the frequency of contemporary gene flow in
multiple linear regression using permutation for significance testing. Note that in all of the above examples, forests are hindering gene flow.

The approach is intuitively appealing in animals, but less
so in plants because pollen and seed dispersal mechanisms
depend on (multiple) biotic or abiotic vectors. In addition,
genetic distances in plants do not account for differences in
seed and pollen dispersal [16].
Given these shortcomings of the landscape distance/
resistance approach in plants, it is not surprising that it
has mainly been applied in special situations. Mantel and
partial Mantel tests of genetic distances with geographic
distances along coasts or streams have been used in studies
on wild sugar beet (Beta vulgaris ssp. maritima [27]) or
Japanese primrose (Primula sieboldii [28]). Similarly, the
effect of differences in flowering time among individuals on
genetic distances has been studied with partial Mantel
tests in several alpine snowbed plants [29,30]. In fact,
researchers can still profit from the landscape distance/
resistance approach in plants when analyzing the influ-
ence of major landscape elements on gene flow in a general
way (Box 1). For instance, forests might act as barriers to
gene flow in a wind-pollinated and wind-dispersed meadow
herb. The occurrence of large forests between populations
or individuals should generally exhibit a negative effect on
gene flow and effectively increase genetic distances.
Researchers could also simultaneously determine the
effects of several landscape distances between pairs of
sampling locations (including geographic distance) on ge-
netic distances in multiple linear regression with permu-
tation-based significance testing using software such as
PERMUTE [31] or BLOSSOM [32]. When testing several
models of landscape distance/resistance, significance
values should be adjusted because of multiple testing,
and model performance has to be evaluated (Box 2) [33],
which has rarely been done in landscape genetics (but see
[34,35]).
Overlay technique
The overlay approach of landscape genetics [7] identifies
population groups, barriers, genetic discontinuities or iso-
lines. These genetic structures are overlaid onto maps of
selected landscape elements such as topography or land

677
Review
Box 1. Five current hot topics in landscape genetics of
plants
(i) Perform landscape distance/resistance analysis to detect the
influence of major landscape elements on gene flow by seed
and pollen, that is correlate genetic distance with landscape
elements (e.g. forests, mountain ridges) or abiotic factors (e.g.
wind direction), and make use of statistical methods alternative
to (partial) Mantel tests.
(ii) Expand the use of overlays in plants and incorporate boundary
overlap statistics.
(iii) Combine estimates of contemporary gene flow by pollen or
seed with landscape data in multiple regression analysis with
permutation testing.
(iv) Evaluate the influence of spatial genetic structure and popula-
tion history in outlier detection or allele distribution analyses.
(v) Describe and prove molecular function of identified outlier
genomic region and provide empirical tests of the selective
relevance of identified adaptive genetic markers in plants (e.g.
transplant experiments).
cover/land use to search for geographical coincidences of
group boundaries, barriers, genetic discontinuities or iso-
lines with landscape elements (Figure 1b).
Various statistical approaches can be used to form
genetic groups of populations or individuals to be used
in overlays. Bayesian clustering [36] based on Hardy-
Weinberg and linkage equilibrium as implemented in
STRUCTURE [37], BAPS-5 [38], TESS [39] or GENE-
LAND [40] are widely used for this purpose. These pro-
grams can also consider coordinates of sampling locations
[38–40]. Alternatively, non-Bayesian S-AMOVA can be
applied to form population groups. This method maximizes
genetic differentiation (FCT) among groups [41]. For in-
stance, no or only weak clustering was detected in the
alpine blue thistle (Eryngium alpinum) [42] and the rain-
forest Anguama tree (Aucoumea klaineana) [43] despite
spatial distribution gaps, and grouping according to river
catchments was inferred in the Chinese maidenhair fern
(Adiantum reniforme var. sinensis) [44]. Other methods
search for areas of strong changes in allele frequencies,
such as the Monmonier algorithm implemented in BAR-
RIER [45] or ALLELES IN SPACE [46] to detect genetic
barriers among populations (e.g. [26]). Individual- or pop-
ulation-based wombling also identifies genetic discontinu-
ities [1,47,48]. Finally, interpolation such as kriging from
principal component analysis (PCA) axis loadings deter-
mines genetic isolines similar to contour lines in topo-
graphical maps (Figure 1b). For instance, kriging helped
in visualizing the small-scale genetic structure in snap-
dragon (Antirrhinum microphyllum) [49]. Membership
coefficients estimated from Bayesian clustering methods
can also be interpolated, providing clustering surface maps
(e.g. [25]).
The major drawback of the common overlay approach
is that spatial coincidence of landscape elements with
genetic discontinuities, barriers or isolines is simply based
on subjective visual inspection. No statistical procedure
is usually involved in this step. Overlays are thus of
exploratory nature and prone to false inference. However,
subjectivity in analysis could be avoided by applying
boundary overlap statistics [50,51]. Such statistics have,
for example, been used to study the spatial coincidence
678
Trends in Plant Science Vol.15 No.12
Box 2. Landscape genetic approach to identify molecular
markers bearing the signature of natural selection
Landscape genetics tries to identify molecular markers whose
changes in allele frequencies are correlated with environmental
factors potentially acting as selective pressures and enforcing
directional natural selection (Figure 2a). Usually, many samples
are taken along environmental gradients or in environmentally
heterogeneous situations, and large genome scans with hundreds
to thousands of co-dominant or dominant markers (AFLPs or SNPs
[16]) are performed. Finally, allele presence/absence (individual-
based analysis) or allele frequencies (population-based analysis) are
correlated with environmental variables taken from geo-referenced
databases or from field surveys [10]. Markers significantly correlated
with environmental factors are considered to be linked to or to be
located within genomic regions under selection, whereas uncorre-
lated molecular markers are considered as neutral, at least with
respect to the particular set of environmental variables tested.
Various statistical methods are used to establish allele distribution
models [9]. First, logistic regression relates allele occurrences with
environmental variables. The spatial analysis method (SAM [87])
offers a user-friendly framework to perform logistic regression in a
landscape genetic approach. SAM has successfully been used in
several animal studies [14,97], but we are aware of only a single
plant study that has applied SAM. Parisod and Joost [98] examined
patterns of selection in populations of buckler mustard (Biscutella
laevigata) characterized by different population histories. Logistic
regression has also been used to correlate allele frequencies at an
outlier locus associated with altitude and temperature in common
beech (Fagus sylvatica) [99]. The logistic regression approach can
easily be extended to more sophisticated generalized linear models
[92]. Recently, we successfully applied multiple linear regression to
identify AFLP fragments correlated with temperature and precipita-
tion in the alpine rock cress (Arabis alpina) [13]. Polynomial
transformation of environmental variables can be included in
multiple linear regression, thus also tracking non-linear adaptive
responses.
The correlative landscape genetic approach to adaptive genetic
variation has several shortcomings. (i) Researchers can only identify
correlations of molecular markers with those environmental factors
that were included in the analysis. Molecular markers not correlated
with these environmental factors are therefore not necessarily
neutral in a general sense. For this reason, researchers might also
wish to apply classical outlier locus detection to their data sets if
population-based sampling is available [97]. (ii) Given that a
potentially large number of statistical tests is applied in explorative
data analysis, it is important to adjust significance values for type I
error inflation due to multiple testing and/or using model evaluation
criteria such as adjusted R2 or the Akaike information criterion (AIC)
values [33]. The use of adjusted R2 and AIC is currently explored in
landscape genetic studies of plants.
between heterogeneity of forest structures and boundaries
of bird territories [52]. To our knowledge, boundary overlap
statistics have rarely been used in landscape genetics
(Box 1).
Despite its appeal for studying sessile organisms, the
overlay technique has not been popular in plants and should
therefore be further explored, especially in connection with
appropriate significance testing. In fact, most plant studies
have dealt with phylogeographic patterns [53], and only a
few have overlaid inferred genetic patterns onto land-cover/
land-use or topographical maps. By contrast, obvious effects
of various landscape elements on gene flow have been
identified with overlays in animal studies (e.g. [54,55]).
Assessment of contemporary gene flow
Landscapes throughout the world are changing at
an unprecedented speed [2]. This poses a problem to
Review
landscape genetics because the landscape distance/resis-
tance and (most) overlay approaches rely on historical
measurements of gene flow (i.e. genetic distance and dif-
ferentiation [15]). However, when landscapes are changing
rapidly, historical gene flow measures tend to reflect the
historical rather than the contemporary landscape [56,57].
In such situations, researchers would like to assess con-
temporary migration and gene flow [6]. To do so, two main
approaches are currently available: parentage analysis
[58] and assignment tests [36] (Figure 1c).
Parentage analysis, and its variants maternity and
paternity analysis, has widely been used in plants, espe-
cially trees [59]. Paternity analysis infers contemporary
pollen flow from the genetic analysis of open-pollinated
offspring of known mothers by using maximum likelihood
methods (e.g. CERVUS [60]) or Bayesian inference (MAS-
TERBAYES [61]) to identify the most probable fathers [58].
Similarly, maternity analysis uses maternal seed coat
tissue [62] or uniparentally inherited organellar DNA
markers (e.g. chloroplast DNA [16]) to detect the most
probable mothers of seedlings or trapped seed. Parentage
analysis asks for complete sampling of all potential parents
in a study area, which logistically limits the study range.
This is a major shortcoming of studies using parentage
analysis, and therefore landscape-scale studies are rare
(but see [63]).
From studies on contemporary gene flow, we have
learned that pollen and seed dispersal are more frequent
and occur over greater distances than expected from eco-
logical investigations [6]. However, the effect of landscapes
has rarely been considered, and most studies simply com-
pared contemporary gene flow patterns in fragmented
versus non-fragmented situations (e.g. [64–66]). Research-
ers have started to assess landscape effects by applying
multiple linear regression with permutation testing of
the frequency of mating events among pairs of individuals
and various landscape elements [31,32] (Figure 1c). We
have also compared realized mating patterns with a
null model of saturated mating among all individuals
studied in a regional population of the service tree (Sorbus
domestica [67]). One problem with the landscape genetic
analysis of data from parentage analysis is that no mating
between particular individuals either reflects a real lack of
mating or simply a low detection probability because of
insufficient sample size in terms of parents and offspring
studied.
Alternatively, contemporary migration and gene flow
are inferred through assignment tests, which are often
used in animals [36]. Assignment tests can discriminate
between first generation migrants, migrants during the
last generation and recent migrants (during the last few
generations) using Bayesian software such as GENE-
CLASS [68] or BAYESASS [69]. Assignment tests on
plants are rare, probably because it is difficult to disentan-
gle gene flow by seed and pollen. However, by using
particular settings in GENECLASS [68], He et al. [70]
restricted assignment to contemporary seed dispersal.
Two major caveats to be considered when using assign-
ment tests are that although assignment tests enable
individual migrants to be identified, migration rates and
populations of origin can only be appropriately identified if
Trends in Plant Science Vol.15 No.12
all populations in a landscape were included in sampling
and analysis.
Several studies have detected that contemporary seed
dispersal occurs over large stretches of unsuitable habitat,
across inhospitable mountain ridges or within river catch-
ments [70–72]. Again, landscape effects have seldom been
combined with assignment tests, although dedicated soft-
ware (BMIR [73]) is available for correlating directional
contemporary migration rates based on Bayesian inference
with landscape distance data in a multiple linear regres-
sion framework. It is evident that additional statistical
tools have to be developed to analyze the full breadth of
contemporary gene flow and migration in a landscape
context (Box 1).
Landscape genetics of adaptive genetic variation
A popular route in landscape genetics of plants has been to
correlate population genetic diversity with environmental
factors at habitat patches. Researchers have studied
whether genetic diversity was related to local soil type,
humidity, vegetation structure or management type (e.g.
[74–76]). These studies were based on the analysis of
neutral markers. However, neutral genetic diversity does
not directly relate to adaptive genetic variation [77,78] and
mainly reflects local population size [79]. Neutral genetic
diversity is only indirectly affected by local environmental
factors if these factors influence processes such as gene flow
or mating [3]. Therefore, if researchers want to establish
the relationship between genetic diversity and environ-
mental factors, they should specifically assess adaptive
genetic variation [2].
The first step in analyzing adaptive genetic variation is
to identify genomic regions bearing signs of selection.
Among various methodological approaches [80], genome
scans have been the preferred method in non-model organ-
isms for about a decade. In this analysis, loci showing a
higher genetic differentiation among populations (FST)
than expected under neutrality (i.e. outlier loci) are iden-
tified out of a large number of loci studied across a genome
[81–89]. This population genomic approach has also been
applied to plants (e.g. [84,85]). Selected environmental
factors are only a posteriori correlated with allele frequen-
cies at outlier loci to infer potential selective pressures [9].
By contrast, the landscape genetic approach [8] directly
uses environmental data to pinpoint molecular markers
linked to or located within genomic regions under selection
[9]. Accordingly, many samples are collected along envi-
ronmental gradients (Figure 2), and large genome scans
with hundreds to thousands of amplified fragment poly-
morphisms (AFLPs) or single-nucleotide polymorphisms
(SNPs) are performed [16,86]. Allele occurrence in indivi-
duals or allele frequency in populations is subsequently
correlated with local environmental conditions, for exam-
ple, estimates of temperature, precipitation, slope, altitude
or habitat type [9,10]. Several statistical methods are used
for this purpose, such as simple linear regression [13] or
generalized linear models, for example the logistic regres-
sion implemented in the spatial analysis method (SAM
[87]) (Box 2). Molecular markers significantly correlated
with environmental factors having substantial effect size
are seen as linked to genomic regions influenced by these
679
 Review
[(Figure_2)TD$IG] Trends in Plant Science Vol.15 No.12

(a) Directional selection (b) Directional selection (c) Directional selection and
on small spatial scale on large spatial scale geographic structure

Allele frequency

Environmental gradient Environmental gradient Environmental gradient

and geographic distance and geographic distance

Unrestricted gene flow Restricted gene flow Restricted gene flow

and population history

TRENDS in Plant Science

Figure 2. Conceptual characteristics of landscape genetic studies assessing adaptive genetic variation. (a) A molecular marker linked to a genomic region under directional
selection shows a change in allele frequency along an environmental gradient (solid red line). By contrast, neutral molecular markers show no such change on a small-
spatial scale (hence, unrestricted gene flow; broken lines) because they are not affected by natural selection (modified from [96]). (b) If directional selection occurs on a
larger spatial scale, geographic distance and environmental gradients often co-vary. In consequence, allele frequencies at neutral loci will randomly change with distance
and, indirectly, along the environmental gradient because of restricted gene flow and genetic drift (i.e. isolation by distance [7]). Therefore, landscape genetic studies on
adaptive genetic variation might falsely identify some neutral markers as bearing the signature of adaptive evolution, that is, spatial genetic structure is a nuisance factor in
analysis (Box 3). (c) The effect presented in (b) is most prominent if spatial genetic structure due to phylogeography or population history has lead to strong changes in
allele frequencies. Here, changes in allele frequencies of markers showing signs of adaptive evolution are expected to be more pronounced than those of neutral loci.

factors. After molecular characterization, identified mar-
kers can be used for cross-validation in other landscapes
or in experiments aiming at verifying their adaptive or
molecular functionality (Boxes 1 and 2) [2,9]. Plants are
particularly more amenable to corresponding experimen-
tation than (most) animals. In the model plant thale cress
(Arabidopsis thaliana), a large SNP genome scan and an in
detail genomic analysis found that one particular allele at
locus ACD6 underpins resistance to microbial infection
and herbivory in natural populations and therefore pro-
vides large fitness advantages under high pathogen and
herbivory pressure, despite severely reducing vegetative
growth [88].
Only a few landscape genetic studies on the adaptive
genetic variation of plants are available so far (Box 2), and
the methodological foundations of the field are not yet fully
explored. In particular, researchers have only started to
deal with the problem of spatial genetic structure [10,89],
potentially interfering with landscape genetic analysis and
leading to the detection of molecular markers falsely found
to be linked to genomic regions under selection (Boxes 1
and 3). Likewise, most studies still lack any indication of
the functionality of the outlier loci identified, be it based on
experimental or molecular evidence.
Perspectives
Landscape genetics of plants is a largely under-explored
field. One reason for this is that applying the classical
landscape distance/resistance approach to infer landscape
effects on gene flow is less amenable to studies of plants
than to animals, owing to the particular means of gene
dispersal through pollen and seed. Also the overlay tech-
nique has rarely been used in plants in a true landscape
genetic setting with sound statistical analysis. Many stud-
ies on contemporary gene flow were small-scale and did not
consider landscape effects. Therefore, significant progress
can be achieved by applying existing methodology to plants
on adequate spatial scales. However, it is apparent that the
development of new statistical tools is necessary to analyze
genetic data in concert with landscape and environmental
data [90]. Interesting recent advances include the incorpo-
ration of graph theory in landscape genetics [91], which
Box 3. Spatial genetic structure as a nuisance factor in
landscape genetic studies on adaptive genetic variation
Spatial genetic structure caused by (i) restricted gene flow and
leading to isolation by distance patterns (Figure 2b) or (ii)
phylogeographic history, range expansion or population demogra-
phy (e.g. bottlenecks; Figure 2c) might substantially interfere with
both population genomic outlier analysis and landscape genetic
approaches. Pronounced genetic structure is thus a nuisance
parameter in landscape genetics of adaptive genetic variation
[9,89]. Excoffier et al. [89] showed that ignoring hierarchical spatial
genetic structure in classical outlier detection analysis results in the
identification of numerous false outlier loci. This study highlights
the need for adequate treatment of spatial genetic structure when
searching for molecular markers linked to genes under selection.
The consideration of spatial genetic structure in landscape genetics
of adaptive genetic variation has only just begun. So far, researchers
have tended to apply Bayesian clustering to their samples and then
use landscape genetic analysis with logistic or linear regression
[13,84] within each genetic cluster separately. Recently, more
sophisticated approaches to account for spatial genetic structure
have been introduced. For instance, mixed linear models allow
controlling for population structure when populations are known
[100]. Accordingly, we used generalized estimating equations (GEE),
which take small-scale autocorrelation of samples into consideration
[92], and we [13] applied principal coordinates of neighbor matrices
(PCNMs [101]) to landscape genetic analysis using R [102]. PCNM
values on large- and small-spatial scales can be introduced as
additional factors in linear regression and account for the effects of
different spatial scales and for the effects of un-accounted environ-
mental factors. Despite such promising new tools, it is obvious that
the issue of spatial genetic structure in landscape genomic research
needs more attention in future analyses as well as the development of
appropriate statistical methods.

680
Review
considers landscape and environmental data at and be-
tween sampling locations. Various types of hierarchical
Bayesian models are also gaining increasing popularity in
ecology and genetics [11]. A conceptual shortcoming of
many landscape genetic studies is the virtual lack of
replication at the landscape level (but see [75,92]) and of
multi-species studies (but see [91,93]). Replicated land-
scape genetic analyses require particularly large genetic
sample sizes and hence should profit from the current
exponential increase in sequencing and genotyping capac-
ity [94]. To deal with the corresponding huge genetic data
sets, bioinformatics already relies on machine learning
techniques. These techniques enable the screening of large
numbers of genetic markers such as SNPs, making them
particularly relevant in genomics and the study of adaptive
genetic variation [95].
Global change, that is the world-wide alteration of
natural and traditionally used landscapes and the rapidly
changing climate, demands profound knowledge about the
migration ability of species as well as their potential to
adapt to new, human-altered environments. Landscape
genetics is ideally suited to provide such relevant real-
world data [2,9]. For this task to be achieved, researchers
have to make full use of existing and newly developed
methodological landscape genetic approaches, especially
so in plants.
Disclosure statement
The authors declare no conflict of interest.
Acknowledgements
R.H., D.B. and F.G. thank the CCES ENHANCE and BIOCHANGE
projects of the ETH domain, the AVE project of the Swiss National
Science Foundation (CRSI33 127155/1) and the European ECOCHANGE
project (FP6-036866) for financial support. S.M. was supported by the
Institut Universitaire de France as a junior member. R.H. and S.M. also
acknowledge support by the National Center for Ecological Analysis and
Synthesis, funded by NSF (Grant DEB-0553768), the University of
California at Santa Barbara, and the State of California. We also thank
an anonymous referee for valuable comments on the manuscript.
References
1 Manel, S. et al. (2003) Landscape genetics: combining landscape
ecology and population genetics. Trends Ecol. Evol. 18, 189–197
2 Reusch, T.B.H. and Wood, T.E. (2007) Molecular ecology of global
change. Mol. Ecol. 16, 3973–3992
3 Ouburg, N.J. et al. (2010) An essay on the necessity and feasibility of
conservation genomics. Conserv. Genet. 11, 643–653
4 Holderegger, R. and Wagner, H.H. (2006) A brief guide to landscape
genetics. Landsc. Ecol. 21, 793–796
5 Storfer, A. et al. (2006) Putting the ‘‘landscape’’ in landscape genetics.
Heredity 98, 128–142
6 Holderegger, R. and Wagner, H.H. (2008) Landscape genetics.
BioScience 58, 199–207
7 Storfer, A. et al. (2010) Landscape genetics: where are we now? Mol.
Ecol. 19, 3496–3514
8 Lowry, D.B. (2010) Landscape evolutionary genomics. Biol. Lett. 6,
502–504
9 Holderegger, R. et al. (2008) Land ahead: using genome scans to
identify molecular markers of adaptive relevance. Plant Ecol. Div.
1, 273–283
10 Manel, S. et al. (2010) Perspectives on the use of landscape genetics to
detect adaptive genetic variation in the field. Mol. Ecol. 19, 3760–3772
11 Segelbacher, G. et al. (2010) Landscape genetics: concepts and
challenges in a conservation context. Conserv. Genet. 11, 375–385
12 Holderegger, R. and Di Giulio, M. (2010) The genetic effects of roads: a
review of empirical evidence. Basic Appl. Ecol. 11, 522–531
Trends in Plant Science Vol.15 No.12
13 Manel, S. et al. (2010) Common factors drive genetic variation of
adaptive relevance on different spatial scales in Arabis alpina. Mol.
Ecol. 19, 3824–3835
14 Joost, S. et al. (2007) A spatial analysis method (SAM) to detect
candidate loci for selection: towards a landscape genomics
approach to adaptation. Mol. Ecol. 16, 3955–3969
15 Whitlock, M.C. and McCauley, D.E. (1999) Indirect measures of gene
flow and migration: FST 6¼ 1/(4Nm+1). Heredity 82, 117–125
16 Lowe, A.J. et al. (2004) Ecological Genetics. Design, Analysis and
Application, Blackwell
17 Adriaensen, F. et al. (2003) The application of ‘‘least-cost’’ modeling as
a functional landscape model. Landsc. Urban Plan. 64, 233–247
18 Baguette, M. and Van Dyck, H. (2007) Landscape connectivity and
animal behaviour: functional grain as a key determinant for dispersal.
Landsc. Ecol. 22, 117–1129
19 Smouse, P.E. et al. (1986) Multiple regression and correlation
extensions of the Mantel tests of matrix correspondence. Syst. Zool.
35, 627–632
20 Wright, S. (1943) Isolation by distance. Genetics 28, 114–138
21 Cushman, S.A. et al. (2006) Gene flow in complex landscapes: testing
multiple hypotheses with causal modeling. Am. Nat. 168, 486–499
22 Raufaste, N. and Rousset, F. (2001) Are partial Mantel tests
adequate? Evolution 55, 1703–1705
23 Balkenhol, N. et al. (2009) Statistical approaches in landscape
genetics: an evaluation of methods for linking landscape and
genetic data. Ecography 32, 818–830
24 Legendre, P. and Fortin, M-J. (2010) Comparison of the Mantel test
and alternative approaches for detecting complex multivariate
relationships in the spatial analysis of genetic data. Mol. Ecol. Res.
10, 831–844
25 McRae, B.H. (2006) Isolation by resistance. Evolution 60, 1551–1561
26 Trénel, P. et al. (2008) Landscape genetics, historical isolation and
cross-Andean gene flow in the wax palm, Ceroxylon echinulatum
(Arecaceae). Mol. Ecol. 17, 3528–3540
27 Fievet, V. et al. (2007) Spatial analysis of nuclear and cytoplasmic DNA
diversity in wild sea beet (Beta vulgaris ssp. maritima) populations: do
marine currents shape genetic structure? Mol. Ecol. 16, 1847–1864
28 Kitamoto, N. et al. (2005) Spatial genetic structure among and within
populations of Primula sieboldii growing on separate streams. Mol.
Ecol. 14, 149–157
29 Hirao, A.S. and Kudo, G. (2004) Landscape genetics of alpine-snowbed
plants: comparisons along geographic and snowmelt gradients.
Heredity 93, 290–298
30 Hirao, A.S. and Kudo, G. (2008) The effect of segregation of flowering
time on fine-scale spatial genetic structure in an alpine-snowbed herb
Primula cuneifolia. Heredity 100, 424–430
31 Legendre, P. et al. (1994) Modeling brain evolution from behaviour: a
permutational regression approach. Evolution 48, 1487–1499
32 Cade, B.S. and Richards, J.D. (2005) User Manual BLOSSOM
Statistical Software, US Geological Survey
33 Johnson, J.B. et al. (2004) Model selection in ecology and evolution.
Trends Ecol. Evol. 19, 101–108
34 Parisod, C. and Christin, P-A. (2008) Genome-wide association to fine-
scale ecological heterogeneity within a continuous population of
Biscutella laevigata (Brassicaceae). New Phytol. 178, 436–447
35 Helm, A. et al. (2009) Human influence lowers plant genetic diversity
in communities with extinction debt. J. Ecol. 97, 1329–1336
36 Manel, S. et al. (2005) Assignment methods: matching biological
questions with appropriate techniques. Trends Ecol. Evol. 20, 136–142
37 Pritchard, J.K. et al. (2000) Inference of population structure using
multilocus genotype data. Genetics 155, 945–959
38 Corander, J. et al. (2008) Enhanced Bayesian modeling in BAPS
software for learning genetic structures of populations. BMC
Bioinformatics 9, 539
39 François, O. et al. (2006) Bayesian clustering using hidden Markov
random fields in spatial population genetics. Genetics 174, 805–816
40 Guillot, G. et al. (2005) GENELAND: a computer package for
landscape genetics. Mol. Ecol. Notes 5, 712–715
41 Dupanloup, I. et al. (2002) A simulated annealing approach to define
the genetic structure of populations. Mol. Ecol. 11, 2571–2581
42 Gaudeul, M. and Till-Bottraud, I. (2008) Genetic structure of the
endangered perennial plant Eryngium alpinum (Apiaceae) in an
alpine valley. Biol. J. Linn. Soc. 93, 667–677
681
Review
43 Born, C. et al. (2008) Small-scale spatial genetic structure in the
Central African rainforest tree species Aucomea klaineana: a stepwise
approach to infer the impact of limited gene dispersal, population
history and habitat fragmentation. Mol. Ecol. 17, 2041–2050
44 Kang, M. et al. (2008) Understanding population structure and
historical demography in a conservation context: population
genetics of an endangered fern. Divers. Distrib. 14, 799–807
45 Manni, F. et al. (2004) Geographic patterns of (genetic, morphologic,
linguistic) variation: how barriers can be detected by using
Monmonier’s algorithm. Hum. Biol. 76, 173–190
46 Miller, M.P. (2005) ALLELES IN SPACE (AIS): computer software for
the joint analysis of interindividual spatial and genetic information.
J. Hered. 96, 722–724
47 Manel, S. et al. (2007) A new individual-based spatial approach for
identifying genetic discontinuities in natural populations. Mol. Ecol.
16, 2031–2043
48 Soares, T.N. et al. (2008) Landscape conservation genetics of Dipetryx
alata (‘‘baru’’ tree: Fabaceae) from Cerrado region of Central Brazil.
Genetica 132, 9–19
49 Torres, E. et al. (2003) Analysis of within-population spatial genetic
structure in Antirrhinum microphyllum (Scrophulariaceae). Am. J.
Bot. 90, 1688–1695
50 Fortin, M-J. and Dale, M.R.T. (2005) Spatial Analysis. A Guide for
Ecologists, Cambridge University Press
51 Fortin, M-J. et al. (2005) Species’ geographic ranges and distribution
limits: pattern analysis and statistical issues. Oikos 108, 7–17
52 St. Louis, V. et al. (2004) Spatial association between forest
heterogeneity and breeding territory boundaries of two forest
songbirds. Landsc. Ecol. 19, 591–601
53 Avise, J.C. (1994) Molecular Markers, Natural History and Evolution,
Chapman and Hall
54 Coulon, A. et al. (2006) Genetic structure is influenced by landscape
features: empirical evidence from a roe deer population. Mol. Ecol. 15,
1669–1679
55 Kuehn, R. et al. (2007) Genetic effect of transportation infrastructure
on roe deer populations (Capreolus capreolus). J. Hered. 98, 13–
22
56 Balkenhol, N. et al. (2009) Identifying future research needs in
landscape genetics: where to from here? Landsc. Ecol. 24, 455–463
57 Anderson, C.D. et al. (2010) Considering spatial and temporal scale in
landscape-genetic studies of gene flow. Mol. Ecol. 19, 3565–3575
58 Jones, A.G. et al. (2010) A practical guide to methods of parentage
analysis. Mol. Ecol. Res. 10, 6–30
59 Sork, V.L. and Smouse, P.E. (2006) Genetic analysis of landscape
connectivity in tree populations. Landsc. Ecol. 21, 821–836
60 Marshall, T.C. et al. (1998) Statistical confidence for likelihood-
based paternity inference in natural populations. Mol. Ecol. 7,
639–655
61 Walling, C.A. et al. (2010) Comparing parentage inference software:
reanalysis of a red deer pedigree. Mol. Ecol. 19, 1914–1928
62 Godoy, J.A. and Jordano, P. (2001) Seed dispersal by animals: exact
identification of source trees with endocarp DNA microsatellites. Mol.
Ecol. 10, 2275–2283
63 Kamm, U. et al. (2009) Frequent long-distance gene flow in a rare
temperate forest tree (Sorbus domestica) on the landscape scale.
Heredity 103, 476–482
64 White, G.M. et al. (2002) Increased pollen flow counteracts
fragmentation in a tropical dry forest: an example from Swietenia
humilis Zuccarini. Proc. Natl. Acad. Sci. U. S. A. 99, 2038–2042
65 Bacles, C.F.E. and Ennos, R.A. (2008) Paternity analysis of pollen-
mediated gene flow for Fraxinus excelsior L. in a chronically
fragmented landscape. Heredity 101, 368–380
66 Hoebee, S.E. et al. (2007) Mating patterns and contemporary gene
flow by pollen in a large continuous and a small isolated population of
the scattered forest tree Sorbus torminalis. Heredity 99, 47–55
67 Kamm, U. et al. (2010) Open areas in a landscape enhance pollen-
mediated gene flow of a tree species: evidence from northern
Switzerland. Landsc. Ecol. 25, 903–911
68 Piry, S. et al. (2004) GENECLASS2: a software for genetic assignment
and first-generation migrant detection. J. Hered. 95, 536–539
69 Wilson, G.A. and Rannala, B. (2003) Bayesian inference of
recent migration rates using multilocus genotypes. Genetics 163,
1177–1191
682
Trends in Plant Science Vol.15 No.12
70 He, T. et al. (2009) Long-distance dispersal of seeds in the fire-tolerant
shrub Banksia attenuata. Ecography 32, 571–580
71 Raffl, C. et al. (2008) Patterns in genetic diversity of Trifolium
pallescens populations do not reflect chronosequence on alpine
glacier forelands. Heredity 100, 526–532
72 Mitsiu, Y. et al. (2010) Multiple spatial scale patterns of genetic
diversity in riparian populations of Ainslianaea faurieana
(Asteraceae) on Yakushima Island, Japan. Am. J. Bot. 97,
101–110
73 Faubet, P. and Gaggiotti, O.E. (2008) A new Bayesian method to
identify the environmental factors that influence recent migration.
Genetics 178, 1491–1504
74 Honnay, O. et al. (2006) Low impact of present and historical
landscape configuration on the genetics of fragmented Anthyllis
vulneraria populations. Biol. Conserv. 127, 411–419
75 Schmidt, T. et al. (2008) Effects of landscape structure on genetic
diversity of Geum urbanum L. populations in agricultural landscapes.
Flora 204, 549–559
76 Vandepitte, K. et al. (2007) Landscape genetics of the self-compatible
forest herb Geum urbanum: effects of habitat age, fragmentation and
local environment. Mol. Ecol. 16, 4171–4179
77 Merilä, J. and Crnokrak, P. (2001) Comparison of genetic
differentiation at marker loci and quantitative traits. J. Evol. Biol.
14, 892–903
78 Reed, D.H. and Frankham, R. (2001) How closely correlated are
molecular and quantitative measures of genetic variation? A meta-
analysis. Evolution 55, 1095–1103
79 Leimu, R. et al. (2006) How general are positive relationships between
plant population size, fitness and genetic variation? J. Ecol. 94,
942–952
80 Hoffmann, A.A. and Willi, Y. (2008) Detecting genetic responses to
environmental change. Nat. Rev. Genet. 9, 421–432
81 Storz, J.F. (2005) Using genome scans of DNA polymorphism to infer
adaptive population divergence. Mol. Ecol. 14, 671–688
82 Stinchcombe, J.R. and Hoekstra, H.E. (2008) Combining population
genomics and quantitative genetics: finding the genes underlying
ecologically important traits. Heredity 100, 158–170
83 Beaumont, M.A. and Balding, D.J. (2004) Identifying adaptive genetic
divergence among populations from genome scans. Mol. Ecol. 13,
969–980
84 Savolainen, V. et al. (2006) Sympatric speciation in palms on an
oceanic island. Nature 441, 210–213
85 Bonin, A. et al. (2007) Population adaptive index: a new method to
help measure intraspecific genetic diversity and prioritize
populations for conservation. Conserv. Biol. 21, 697–708
86 Meudt, H.M. and Clarke, A.C. (2007) Almost forgotten or latest
practice? AFLP applications, analyses and advances. Trends Plant
Sci. 12, 106–117
87 Joost, S. et al. (2008) Spatial analysis method (SAM): a software tool
combining molecular and environmental data to identify candidate
loci for selection. Mol. Ecol. Res. 8, 957–960
88 Todesco, M. et al. (2010) Natural allelic variation underlying a
major fitness trade-off in Arabidopsis thaliana. Nature 465,
632–636
89 Excoffier, L. et al. (2009) Detecting loci under selection in a
hierarchically structured population. Heredity 103, 285–298
90 Guillot, G. et al. (2009) Statistical methods in spatial genetics. Mol.
Ecol. 18, 4734–4756
91 Fortuna, M.A. et al. (2009) Networks of spatial genetic variation
across species. Proc. Natl. Acad. Sci. U. S. A. 106, 19044–19049
92 Poncet, B.N. et al. (2010) Tracking genes of ecological relevance using
a genome scan in two independent regional population samples of
Arabis alpina. Mol. Ecol. 19, 2896–2907
93 Alvarez, N. et al. (2009) History or ecology? Substrate type as a major
driver of spatial genetic structure in alpine plants. Ecol. Lett. 12,
632–640
94 Nordborg, M. and Weigel, D. (2008) Next-generation genetics in
plants. Nature 456, 720–723
95 Besenbacher, S. et al. (2009) A fast algorithm for genome-wide
haplotype pattern mining. BMC Bioinformatics 10, S74
96 Schmidt, P.S. et al. (2008) Ecological genetics in the North Atlantic:
environmental gradients and adaptation at specific loci. Ecology 89,
S91–S107
Review Trends in Plant Science Vol.15 No.12

97 Pariset, L. et al. (2009) Landscape genomics and biased FST
approaches reveal single nucleotide polymorphisms under selection
in goat breeds of North-East Mediterranean. BMC Genetics 10, 7
98 Parisod, C. and Joost, S. (2010) Divergent selection in trailing- versus
leading-edge populations of Biscutella laevigata. Ann. Bot. 105, 655–660
99 Jump, A.S. et al. (2006) Natural selection and climate change:
temperature-linked spatial and temporal trends in gene frequency
in Fagus sylvatica. Mol. Ecol. 15, 3469–3480
100 Bradbury, P.J. et al. (2007) TASSEL: software for association
mapping of complex traits in diverse samples. Bioinformatics 23,
2633–2635
101 Borcard, D. and Legendre, P. (2002) All-scale spatial analysis of
ecological data by means of principal coordinates of neighbour
matrices. Ecol. Model. 153, 51–68
102 R Development Core Team (2007) R: a Language and Environment for
Statistical Computing (http://www.R-project.org)

Plant Science Conferences in 2011

5th Biennial Conference of the International Biogeography Society
7–11 January, 2011
Herakleon, Greece
http://www.biogeography.org/html/Meetings/index.html

Plant and Animal Genome XIX Conference

15–19 January, 2011
San Diego, USA
http://www.intl-pag.org/

Plant Abiotic Stress Tolerance Mechanisms, Water and Global Agriculture

17–22 January, 2011
Keystone, USA
http://www.keystonesymposia.org/Meetings/

---------------------------------------------------------------------------------------------------------------------------------------------------------------------------
Suggest a conference
Please use the form at http://www.cell.com/conferences/SuggestConference to suggest a
conference for Cell Press the Conference Calendar.

683